rmontanana/BayesNet
1
0
Fork 0
Code Issues 2 Pull Requests 1 Actions Packages Projects Releases 6 Wiki Activity

Fix weights mistakes in computation

Browse Source
This commit is contained in:
Ricardo Montañana Gómez
2023-08-16 12:32:51 +02:00
parent 4d4780c1d5
commit 80b20f35b4
16 changed files with 262 additions and 75 deletions
Download Patch File Download Diff File Expand all files Collapse all files

12
src/BayesNet/BayesMetrics.cc
View File

@@ -38,12 +38,14 @@ namespace bayesnet {
auto source = vector<string>(features);
source.push_back(className);
auto combinations = doCombinations(source);
double totalWeight = weights.sum().item<double>();
// Compute class prior
auto margin = torch::zeros({ classNumStates });
auto margin = torch::zeros({ classNumStates }, torch::kFloat);
for (int value = 0; value < classNumStates; ++value) {
auto mask = samples.index({ -1, "..." }) == value;
margin[value] = mask.sum().item<float>() / samples.size(1);
margin[value] = mask.sum().item<double>() / samples.size(1);
}
cout << "Margin: " << margin;
for (auto [first, second] : combinations) {
int index_first = find(features.begin(), features.end(), first) - features.begin();
int index_second = find(features.begin(), features.end(), second) - features.begin();
@@ -54,7 +56,7 @@ namespace bayesnet {
auto second_dataset = samples.index({ index_second, mask });
auto weights_dataset = weights.index({ mask });
auto mi = mutualInformation(first_dataset, second_dataset, weights_dataset);
auto pb = margin[value].item<float>();
auto pb = margin[value].item<double>();
accumulated += pb * mi;
}
result.push_back(accumulated);
@@ -81,7 +83,7 @@ namespace bayesnet {
double Metrics::entropy(const torch::Tensor& feature, const torch::Tensor& weights)
{
torch::Tensor counts = feature.bincount(weights);
int totalWeight = counts.sum().item<int>();
double totalWeight = counts.sum().item<double>();
torch::Tensor probs = counts.to(torch::kFloat) / totalWeight;
torch::Tensor logProbs = torch::log(probs);
torch::Tensor entropy = -probs * logProbs;
@@ -95,7 +97,7 @@ namespace bayesnet {
unordered_map<int, unordered_map<int, double>> jointCounts;
double totalWeight = 0;
for (auto i = 0; i < numSamples; i++) {
jointCounts[secondFeature[i].item<int>()][firstFeature[i].item<int>()] += 1;
jointCounts[secondFeature[i].item<int>()][firstFeature[i].item<int>()] += weights[i].item<double>();
totalWeight += weights[i].item<float>();
}
if (totalWeight == 0)

22
src/BayesNet/BoostAODE.cc
View File

@@ -1,10 +1,32 @@
#include "BoostAODE.h"
#include "FeatureSelect.h"
namespace bayesnet {
BoostAODE::BoostAODE() : Ensemble() {}
void BoostAODE::buildModel(const torch::Tensor& weights)
{
models.clear();
int n_samples = dataset.size(1);
int n_features = dataset.size(0);
features::samples_t vsamples;
for (auto i = 0; i < n_samples; ++i) {
auto row = dataset.index({ "...", i });
// convert row to std::vector<int>
auto vrow = vector<int>(row.data_ptr<int>(), row.data_ptr<int>() + row.numel());
vsamples.push_back(vrow);
}
auto vweights = features::weights_t(n_samples, 1.0 / n_samples);
auto row = dataset.index({ -1, "..." });
auto yv = features::labels_t(row.data_ptr<int>(), row.data_ptr<int>() + row.numel());
auto featureSelection = features::SelectKBestWeighted(vsamples, yv, vweights, n_features, true);
auto features = featureSelection.fit().getFeatures();
// features = (
// CSelectKBestWeighted(
// self.X_, self.y_, weights, k = self.n_features_in_
// )
// .fit()
// .get_features()
auto scores = features::score_t(n_features, 0.0);
for (int i = 0; i < features.size(); ++i) {
models.push_back(std::make_unique<SPODE>(i));
}

6
src/BayesNet/CMakeLists.txt
View File

@@ -1,7 +1,9 @@
include_directories(${BayesNet_SOURCE_DIR}/lib/mdlp)
include_directories(${BayesNet_SOURCE_DIR}/lib/Files)
include_directories(${BayesNet_SOURCE_DIR}/lib/featureselect)
include_directories(${BayesNet_SOURCE_DIR}/src/BayesNet)
include_directories(${BayesNet_SOURCE_DIR}/src/Platform)
add_library(BayesNet bayesnetUtils.cc Network.cc Node.cc BayesMetrics.cc Classifier.cc
KDB.cc TAN.cc SPODE.cc Ensemble.cc AODE.cc TANLd.cc KDBLd.cc SPODELd.cc AODELd.cc BoostAODE.cc Mst.cc Proposal.cc ${BayesNet_SOURCE_DIR}/src/Platform/Models.cc)
target_link_libraries(BayesNet mdlp ArffFiles "${TORCH_LIBRARIES}")
KDB.cc TAN.cc SPODE.cc Ensemble.cc AODE.cc TANLd.cc KDBLd.cc SPODELd.cc AODELd.cc BoostAODE.cc
Mst.cc Proposal.cc ${BayesNet_SOURCE_DIR}/src/Platform/Models.cc)
target_link_libraries(BayesNet mdlp FeatureSelect "${TORCH_LIBRARIES}")

6
src/BayesNet/Classifier.cc
View File

@@ -43,7 +43,7 @@ namespace bayesnet {
{
dataset = X;
buildDataset(y);
const torch::Tensor weights = torch::ones({ dataset.size(1) }, torch::kFloat);
const torch::Tensor weights = torch::full({ dataset.size(1) }, 1.0 / dataset.size(1), torch::kFloat);
return build(features, className, states, weights);
}
// X is nxm where n is the number of features and m the number of samples
@@ -55,13 +55,13 @@ namespace bayesnet {
}
auto ytmp = torch::tensor(y, kInt32);
buildDataset(ytmp);
const torch::Tensor weights = torch::ones({ dataset.size(1) }, torch::kFloat);
const torch::Tensor weights = torch::full({ dataset.size(1) }, 1.0 / dataset.size(1), torch::kFloat);
return build(features, className, states, weights);
}
Classifier& Classifier::fit(torch::Tensor& dataset, vector<string>& features, string className, map<string, vector<int>>& states)
{
this->dataset = dataset;
const torch::Tensor weights = torch::ones({ dataset.size(1) }, torch::kFloat);
const torch::Tensor weights = torch::full({ dataset.size(1) }, 1.0 / dataset.size(1), torch::kFloat);
return build(features, className, states, weights);
}
Classifier& Classifier::fit(torch::Tensor& dataset, vector<string>& features, string className, map<string, vector<int>>& states, const torch::Tensor& weights)

5
src/BayesNet/Network.cc
View File

@@ -5,7 +5,6 @@
namespace bayesnet {
Network::Network() : features(vector<string>()), className(""), classNumStates(0), fitted(false) {}
Network::Network(float maxT) : features(vector<string>()), className(""), classNumStates(0), maxThreads(maxT), fitted(false) {}
Network::Network(float maxT, int smoothing) : laplaceSmoothing(smoothing), features(vector<string>()), className(""), classNumStates(0), maxThreads(maxT), fitted(false) {}
Network::Network(Network& other) : laplaceSmoothing(other.laplaceSmoothing), features(other.features), className(other.className), classNumStates(other.getClassNumStates()), maxThreads(other.
getmaxThreads()), fitted(other.fitted)
{
@@ -174,6 +173,7 @@ namespace bayesnet {
void Network::completeFit(const map<string, vector<int>>& states, const torch::Tensor& weights)
{
setStates(states);
laplaceSmoothing = 1.0 / samples.size(1); // To use in CPT computation
int maxThreadsRunning = static_cast<int>(std::thread::hardware_concurrency() * maxThreads);
if (maxThreadsRunning < 1) {
maxThreadsRunning = 1;
@@ -347,7 +347,7 @@ namespace bayesnet {
}
// Normalize result
double sum = accumulate(result.begin(), result.end(), 0.0);
transform(result.begin(), result.end(), result.begin(), [sum](double& value) { return value / sum; });
transform(result.begin(), result.end(), result.begin(), [sum](const double& value) { return value / sum; });
return result;
}
vector<string> Network::show() const
@@ -435,6 +435,7 @@ namespace bayesnet {
{
for (auto& node : nodes) {
cout << "* " << node.first << ": (" << node.second->getNumStates() << ") : " << node.second->getCPT().sizes() << endl;
cout << node.second->getCPT() << endl;
}
}
}

3
src/BayesNet/Network.h
View File

@@ -13,7 +13,7 @@ namespace bayesnet {
int classNumStates;
vector<string> features; // Including classname
string className;
int laplaceSmoothing = 1;
double laplaceSmoothing;
torch::Tensor samples; // nxm tensor used to fit the model
bool isCyclic(const std::string&, std::unordered_set<std::string>&, std::unordered_set<std::string>&);
vector<double> predict_sample(const vector<int>&);
@@ -25,7 +25,6 @@ namespace bayesnet {
void setStates(const map<string, vector<int>>&);
public:
Network();
explicit Network(float, int);
explicit Network(float);
explicit Network(Network&);
torch::Tensor& getSamples();

4
src/BayesNet/Node.cc
View File

@@ -84,7 +84,7 @@ namespace bayesnet {
}
return result;
}
void Node::computeCPT(const torch::Tensor& dataset, const vector<string>& features, const int laplaceSmoothing, const torch::Tensor& weights)
void Node::computeCPT(const torch::Tensor& dataset, const vector<string>& features, const double laplaceSmoothing, const torch::Tensor& weights)
{
dimensions.clear();
// Get dimensions of the CPT
@@ -111,7 +111,7 @@ namespace bayesnet {
coordinates.push_back(dataset.index({ parent_index, n_sample }));
}
// Increment the count of the corresponding coordinate
cpTable.index_put_({ coordinates }, cpTable.index({ coordinates }) + weights.index({ n_sample }).item<float>());
cpTable.index_put_({ coordinates }, cpTable.index({ coordinates }) + weights.index({ n_sample }).item<double>());
}
// Normalize the counts
cpTable = cpTable / cpTable.sum(0);

2
src/BayesNet/Node.h
View File

@@ -26,7 +26,7 @@ namespace bayesnet {
vector<Node*>& getParents();
vector<Node*>& getChildren();
torch::Tensor& getCPT();
void computeCPT(const torch::Tensor& dataset, const vector<string>& features, const int laplaceSmoothing, const torch::Tensor& weights);
void computeCPT(const torch::Tensor& dataset, const vector<string>& features, const double laplaceSmoothing, const torch::Tensor& weights);
int getNumStates() const;
void setNumStates(int);
unsigned minFill();

2
src/BayesNet/TAN.cc
View File

@@ -22,6 +22,8 @@ namespace bayesnet {
auto root = mi[mi.size() - 1].first;
// 2. Compute mutual information between each feature and the class
auto weights_matrix = metrics.conditionalEdge(weights);
cout << "*** Weights matrix ***\n";
cout << weights_matrix << "\n";
// 3. Compute the maximum spanning tree
auto mst = metrics.maximumSpanningTree(features, weights_matrix, root);
// 4. Add edges from the maximum spanning tree to the model